CONCAVITY, INFLECTIONS, CUSPS, TANGENTS, AND ASYMPTOTES

MATH 152, SECTION 55 (VIPUL NAIK)

Corresponding material in the book: Section 4.6, 4.7.
Diculty level: Moderate to hard. If you have seen these topics in AP Calculus, then moderate
diculty; if you havent, then hard.
What students should denitely get: The denitions of concave up, concave down, and point of
inection. The strategies to determine limits at innity, limits valued at innity, vertical tangents, cusps,
vertical asymptotes, and horizontal asymptotes.
What students should hopefully get: The intuitive meanings of these concepts, important examples
and boundary cases, the signicance of concavity in determining local extrema, the use of higher derivative
tests. Important tricks for calculating limits at innity.
Executive summary
0.1. Concavity and points of inection. Words ...
(1) A function is called concave up on an interval if it is continuous and its rst derivative is continuous
and increasing on the interval. If the function is twice dierentiable, this is equivalent to requiring
that the second derivative be positive except possibly at isolated points, where it can be zero. (Think
x
4
, whose rst derivative, 4x
3
, is increasing, and the second derivative is positive everywhere except
at 0, where it is zero).
(2) A function is called concave down on an interval if it is continuous and its rst derivative is continuous
and decreasing on the interval. If the function is twice dierentiable, this is equivalent to requiring
that the second derivative be negative except possibly at isolated points, where it can be zero.
(3) A point of inection is a point where the sense of concavity of the function changes. A point of
inection for a function is a point of local extremum for the rst derivative.
(4) Geometrically, at a point of inection, the tangent line to the graph of the function cuts through the
graph.
Actions ...
(1) To determine points of inection, we rst nd critical points for the rst derivative (which are points
where this derivative is zero or undened) and then use the rst or second derivative test at these
points. Note that these derivative tests are applied to the rst derivative, so the rst derivative here
is the second derivative and the second derivative here is the third derivative.
(2) In particular, if the second derivative is zero and the third derivative exists and is nonzero, we have
a point of inection.
(3) A point where the rst two derivatives are zero could be a point of local extremum or a point of
inection. To nd out which one it is, we either use sign changes of the derivatives, or we use higher
derivatives.
(4) Most importantly, the second derivative being zero does not automatically imply that we have a
point of inection.
0.2. Tangents, cusps, and asymptotes. Words...
(1) We say that f has a horizontal asymptote with value L if lim
x
f(x) = L or lim
x
f(x) = L.
Sometimes, both might occur. (In fact, in almost all the examples you have seen, the limits at ,
if nite, are both equal).
(2) We say that f has a vertical asymptote at c if lim
xc
f(x) = and/or lim
xc
+ f(x) = .
Note that in this case, it usually also happens that f

(x) on the relevant side, with the sign

1
the same as that of f(x)s approach if the approach is from the left and opposite to that of f(x)s
approach if the approach is from the right. However, this is not a foregone conclusion.
(3) We say that f has a vertical tangent at the point c if f is continuous (and nite) at c and
lim
xc
f

(x) = , with the same sign of innity from both sides. If f is increasing, this sign
is +, and if f is decreasing, this sign is . Geometrically, points of vertical tangent behave a
lot like points of inection (in the sense that the tangent line cuts through the graph). Think x
1/3
.
(4) We say that f has a vertical cusp at the point c if f is continuous (and nite) at c and lim
xc
f

(x)
and lim
xc
+ f

(x) are innities of opposite sign. In other words, f takes a sharp about-turn at the
x-value of c. Think x
2/3
.
(5) We say that f is asymptotic to g if lim
x
f(x) g(x) = lim
x
f(x) g(x) = 0. In other words,
the graphs of f and g come progressively closer as |x| becomes larger. (We can also talk of one-sided
asymptoticity, i.e., asymptotic only in the positive direction or only in the negative direction). When
g is a nonconstant linear function, we say that f has an oblique asymptote. Horizontal asymptotes
are a special case, where one of the functions is a constant function.
Actions...
(1) Finding the horizontal asymptotes involves computing limits as the domain value goes to innity.
Finding the vertical asymptotes involves locating points in the domain, or the boundary of the
domain, where the function limits o to innity. For both of these, it is useful to remember the
various rules for limits related to innities.
(2) Remember that for a vertical tangent or vertical cusp at a point, it is necessary that the function
be continuous (and take a nite value). So, we not only need to nd the points where the derivative
goes o to innity, we also need to make sure those are points where the function is continuous.
Thus, for the function f(x) = 1/x, f

(x) on both sides as x 0, but we do not obtain a

vertical tangent rather, we obtain a vertical asymptote.
1. Concavity and points of inflection
1.1. Concavity. Concave up means that the derivative of the function (which measures its rate of change)
is itself increasing. Formally, a function f dierentiable on an open interval I is termed concave up on I if
f

is increasing on I. I hope you remember the denition of an increasing function: it means that for two
points x
1
, x
2
I, with x
1
< x
2
, we have f

(x
1
) < f

(x
2
).
Heres three points:
(1) If f itself is increasing (so that f

is positive), then being concave up means that f is increasing at

an increasing rate. In other words, the slope of the tangent line to the graph of f becomes steeper
and steeper (up) as we go from left to right. Heres a typical picture:
2
(2) If f itself is decreasing (so that f

is negative), then being concave up means that f is decreasing

at a decreasing rate. In other words, the slope of the tangent line to the graph is negative, but it is
becoming less and less steep as we go from left to right. Heres a typical picture:
(3) If f is twice dierentiable, i.e., f

is dierentiable, then we can deduce whether f

is increasing by
looking at f

. Specically, if f

is continuous on I, and f

> 0 everywhere on I except at a few

isolated points, then f is concave up throughout.
Similarly, if f is dierentiable on an open interval I, we say that f is concave down on I if f

is decreasing
on the interval I. I hope you remember the denition of a decreasing function: it means that for two points
x
1
, x
2
I, with x
1
< x
2
, we have f

(x
1
) > f

(x
2
).
Heres three points:
(1) If f itself is decreasing (so that f

is negative), then being concave down means that f is decreasing

at an increasing rate. In other words, the slope of the tangent line to the graph of f becomes steeper
and steeper (downward) as we go from left to right.
(2) If f itself is increasing (so that f

is positive), then being concave down means that f is increasing

at a decreasing rate. In other words, the slope of the tangent line to the graph is positive, but it is
becoming less and less steep as we go from left to right.
3
(3) If f is twice dierentiable, i.e., f

is dierentiable, then we can deduce whether f

is increasing by
looking at f

. Specically, if f

is continuous on I, and f

< 0 everywhere on I except at a few

isolated points, then f is concave down throughout.
1.2. Points of inection. A point of inection is a point c in the interior of the domain of a dierentiable
function (i.e., the function is dened and dierentiable on an open interval containing that point) such that
the function is concave in one sense to the immediate left of c and concave in the other sense to the immediate
right of c.
Another way of thinking of this is that points of inection of a function are points where the derivative
is increasing to the immediate left and decreasing to the immediate right, or decreasing to the immediate
left and increasing to the immediate right. In other words, it is a point of local maximum or a point of local
minimum for the derivative of the function.
Recall that earlier, we noted that for a point of local maximum or a point of local minimum, either the
derivative is zero or the derivative does not exist. Since everything were talking about now is related to f

,
we have that for a point of inection, either f

= 0 or f

does not exist.

So the upshot: concave up means the derivative is increasing, concave down means the derivative is
decreasing, point of inection means the sense in which the derivative is changing changes at the point.
1.3. A point of inection where the rst two derivatives are zero. We now consider one kind of
point of inection: where the rst derivative and the second derivative are both zero. Lets begin with the
example.
Consider the function f(x) := x
3
. Recall rst that since f is a cubic function, it has odd degree, so as
x , f(x) , and as x , we also have f(x) . Further, if we compute f

(x), we get 3x
2
.
Note that the function 3x
2
is positive for x = 0, and is 0 at x = 0. So, from our prior discussion of increasing
and decreasing functions, we see that f is increasing on (, 0] and then again on [0, ). And since the
point 0 is common to the two intervals, f is in fact increasing everywhere on (, ).
If you remember, this was an important and somewhat weird example because, although f

(0) = 0, f
does not attain a local extreme value at 0. This is because the derivative of f is positive on both sides of 0.
4
This was the picture we had from our earlier analysis. But now, with the concepts of concave up, concave
down, and points of inection, we can get a better understanding of whats going on. Specically, we see
that the second derivative f

is 6x, which is negative for x < 0, zero for x = 0, and positive for x > 0. Thus,
f is concave down for x < 0, x = 0 is a point of inection, and f is concave up for x > 0.
So, heres the picture: for x < 0, f is negative, f

is positive, and f

is again negative. Thus, the graph

of f is below the x-axis (approaching 0), it is going upward, and it is going up at a decreasing rate. So, as
x 0, the graph becomes atter and atter.
For x > 0, f is positive, f

is positive, and f

is again positive. Thus, the graph of f is above the x-axis

(starting from the origin), it is going upward, and it is going up at an increasing rate. The graph starts out
from at and becomes steeper and steeper.
So, x = 0 is a no sign change point for f

, which is why it is not a point of local maximum or local

minimum. This is because on both sides of 0, f

is positive. What happens is that it is going down from

positive to zero and then up again from zero to positive. But on a related note, because f

itself dips down

to zero and then goes back up, the point 0 is a point of local minimum for f

, so it is a point of inection
for f.
The main thing you should remember is that when we have a critical point for a function, where the
derivative is zero, but it is neither a point of local maximum nor a point of local minimum, then it is likely
to be a point of inection. In other words, this idea of something that is increasing (or decreasing) and
momentarily stops in its tracks, is the picture of neither a local maximum nor a local minimum but a point
of inection.
1.4. Points of inection where the derivative is not zero.
Lets review the graph of the sine function. The sine function starts with sin(0) = 0, goes up from 0 to
/2, where it reaches the value 1, then drops down to 0, drops down further to 1 at 3/2, and then turns
back up to reach 0 at 2. And this pattern repeats periodically.
5
So far, youve taken me on faith about the way the graph curves. But we can now start looking at things
in terms of concave up and concave down.
The derivative of the sin function is the cos function. Lets graph the cos function. This starts with the
value 1 at x = 0, goes down to zero at x = /2, dips down to 1 at x = , goes back up to 0 at x = 3/2,
and then goes up to 1 at x = 2.
We see that cos is positive from 0 to /2, and sin is increasing on that interval. cos is negative from
/2 to 3/2, and sin is decreasing on that interval. cos is again positive from 3/2 to 2, and sin is again
increasing on that interval.
Next, we want to know where sin is concave up and where it is concave down. And for this, we look at
the second derivative of sin, which is the function sin. As you know, the graph of this function is the same
as the graph of sin, but ipped about the x-axis. This means that where sin is positive, its second derivative
is negative, and where sin is negative, its second derivative is positive.
So, from the interval between 0 and , sin is concave down and on the interval between and 2, sin is
concave up. Breaking the interval down further, sin is increasing and concave down on (0, /2), decreasing
and concave down on (/2, ), decreasing and concave up on (, 3/2), and increasing and concave up on
(3/2, 2). The behavior repeats periodically.
Now, lets concentrate on the points of inection. Note that the sense of concavity changes at multiples
of at the point 0, the function changes from concave up to concave down. At the point , the function
changes from concave down to concave up. Another way of thinking about this is that just before , the
function is decreasing at an increasing rate it is becoming progressively steeper. But from onwards, it
starts decreasing at a decreasing rate, in the sense that it starts becoming less steep. So is the point where
the way the tangent line is turning starts changing.
1.5. A graphical characterization of inection points. Inection points are graphically special because
they are points where the way the tangent line is turning changes sense. Theres a related characterization. If
you draw the tangent line through an inection point, the tangent line cuts through the curve. Equivalently,
the curve crosses the tangent line. This is opposed to any other point, where the curve locally lies to one
side of the tangent line.
For instance, for the cube function f(x) := x
3
, the tangent line is the x-axis, and the curve crosses the
x-axis at x = 0. We see something similar for the tangent lines at the points of inection 0 and for the sin
function.
1.6. Third and higher derivatives: exploration. (I may not get time to cover this in class).
A while ago, we had developed criteria to determine whether a critical point is a point where a local
extreme value is attained. We discussed two tests that could be used: the rst derivative test and the second
derivative test. The rst derivative test said that if f

changes sign across the critical point, it is a point

6
where a local extreme value occurs: a local maximum if the sign change is from positive on the left to
negative on the right, and a local minimum if the sign change is from negative on the left to positive on the
right.
The second derivative test was a test specially suited for functions that are twice dierentiable at the
critical point. This test states that if the second derivative at a critical point is negative, the function attains
a local maximum, and if the second derivative is positive, the function attains a local minimum. This leaves
one case open: what happens if the second derivative at the critical point is zero?
In this case, things are inconclusive. We might have a point of local maximum, a point of local minimum,
an inection point, or none of the above. How do we gure this out? I will give two general principles of
alternation, and then we will look at some examples:
(1) If c is a point of inection for f

and and f

(c) = 0, then c is a point of local extremum for f. If the

point of inection is a change from concave up to concave down, we get a local maximum and if the
change is from concave down to concave up, we get a local minimum.
(2) If c is a point of local maximum or minimum for f

, then c is a point of inection for f. Local

maximum implies a change from concave up to concave down and local minimum implies a change
from concave down to concave up.
Lets illustrate this with the function f(x) := x
5
and the point c = 0. Lets also assume you knew nothing
except dierentiation and applying the derivative tests. We have f

(x) = 5x
4
, f

(x) = 20x
3
, f

(x) = 60x
2
,
f
(4)
(x) = 120x, and f
(5)
(x) = 120. At c = 0, f
(5)
is the rst nonzero derivative.
Now, 0 is a point at which f
(4)
= 0 and f
(5)
> 0. Thus, by the second derivative test, 0 is a point of
local minimum for f
(3)
. So, 0 is a point of inection for f
(2)
, by point (2) above. Thus, 0 is a point of local
minimum for f

, by point (1) above. Thus, 0 is a point of inection for f, by point (2) above.
So, the upshot of this is the alternating behavior between derivatives.
1.7. Higher derivative tests. The discussion above gives a practical criterion to simply use evaluation of
derivatives to determine whether a critical point, where a function is innitely dierentiable, is a point of
local maximum, point of local minimum, point of inection, or none of these.
Suppose f is an innitely dierentiable function around a critical point c for f.. Let k be the smallest
integer for which f
(k)
(c) = 0 and let L be the nonzero value of the k
th
derivative. Then:
(1) If k is odd, then c is a point of inection for f and hence neither a point of local maximum nor a
point of local minimum.
(2) If k is even and L > 0, then c is a point of local minimum for f.
(3) If k is even and L < 0, then c is a point of local maximum for f.
For instance, for the power function f(x) := x
n
, n 2. 0 is a critical point and f is innitely dierentiable.
In this case, c = 0, k = n, and L = n! > 0. Thus, if n is even, then f does attain a local minimum at 0. if n
is odd, 0 is a point of inection. In this simple situation, we could have deduced this directly from the rst
derivative test for n even, the rst derivative changes sign fron negative to positive at 0, because n 1 is
odd. For n odd, n 1 is even, so the rst derivative has positive sign on both sides of 0. However, the good
news is that this general method is applicable for other situations where the rst derivative test is harder to
apply.
1.8. Notion of concave up and concave down for one-sided dierentiable. So far, we have dened
the notion of concave up and concave down on an interval assmuing the function is dierentiable everywhere
on the interval. In higher mathematics, a somewhat more general denition is used, and this makes sense
for functions that have one-sided derivatives everywhere.
Note: Please, please, please, please make sure you understand this clearly: we can calculate the left-hand
derivative and right-hand derivative using the formal expressions only after we have checked that the function
is continuous from that side! If there is a piecewise description of the function and it is not continuous from
one side where the denition is changing, then the corresponding one-sided derivative is not, repeat not
dened.
Suppose f is a function on an interval I = (a, b) such that both the left-hand derivative and the right-hand
derivative of f are dened everywhere on I. Note that both one-sided derivatives being dened at every point
in particular means that the function is continuous at each point, and hence on I. However, f need not be
7
dierentiable at every point, because it is possible that the left-hand derivative and right-hand derivative
dier at dierent points.
We then say that:
(1) f is concave up if, at every point, the right-hand derivative is greater than or equal to the left-hand
derivative, and both one-sided derivative are increasing functions on R.
(2) f is concave down if, at every point, the right-hand derivative is less than or equal to the left-hand
derivative, and both one-sided derivatives are decreasing functions on R.
For instance, rst consider the function f on (0, ):
f(x) := {
x
2
, 0 < x 1
x
3
, 1 < x
Before we proceed further, we check/note that the function is continuous at 1. Indeed it is. Hence, to
calculate the left-hand derivative and right-hand derivative at 1, we can formally dierentiate the expressions
at 1. We obtain that the left-hand derivative at 1 is 2 1 = 2 and the right-hand derivative at 1 is 3 1
2
= 3.
We thus obtain the following piecewise denitions for the left-hand derivative and right-hand derivative:
LHD of f at x = {
2x, 0 < x 1
3x
2
, 1 < x
and:
RHD of f at x = {
2x, 0 < x < 1
3x
2
, 1 x
The derivative is undened at 1. Note that both one-sided derivatives are increasing everywhere, and at
the point 1, where the function is not dierentiable, the right-hand derivative is bigger. Thus, the function
is concave up on (0, ).
Heres the graph, with dashed lines indicating the one-sided derivatives:
On the other hand, consider the function g on (0, ):
g(x) := {
x
3
, 0 < x 1
x
2
, 1 < x
8
The function g is continuous and has one-sided derivatives everywhere. Also note that on the intervals
(0, 1) and (1, ), g is concave up. However, at the critical point 1 where g

is undened, the right-hand

derivative is smaller than the left-hand derivative. Thus, the function is not concave up overall on (0, ),
because at the critical point, its rate of increase takes a plunge for the worse. Heres the picture of g:
1.9. Graphical properties of concave functions. Here are some properties of the graphs of functions
that are concave up, which are particularly important in the context of optimization. You should be able to
do suitable role changes and obtain corresponding properties for concave down functions. In all the points
below, f is a continuous function on an interval [a, b] and is concave up on the interior (a, b).
(1) The only possibilities for the increase-decrease behavior of f are: increasing throughout, decreasing
throughout, or decreasing rst and then increasing.
(2) In particular, f either has exactly one local minimum or exactly one endpoint minimum, and this
local or endpoint minimum is also the absolute minimum.
(3) Also, f cannot have a local maximum in its interior. It has exactly one endpoint maximum, and this
is also the absolute maximum.
(4) For any two points x
1
, x
2
in the domain of f, the part of the graph of f between (x
1
, f(x
1
)) and
(x
2
, f(x
2
)) lies below the chord joining the points (x
1
, f(x
1
)) and (x
2
, f(x
2
)).
(5) If we assume that f is dierentiable on (a, b), the tangent line through any point (x, f(x)) for
a < x < b does not intersect the curve at any other point. In the more general notion where f has
one-sided derivatives, both the left and right tangent line satisfy this property.
For concave down functions, the role of minimum and maximum gets interchanged, and in point (3) above,
the graph is now above the chord rather than below.
1.10. Addendum: concave and convex. The book uses the terminology concave up and concave down,
but its worth knowing that in much of mathematics as well as applications of mathematics, the term convex
is used for concave up and the term concave is used for concave down. However, there is some confusion
about this since some people use concave for concave up and convex for concave down.
9
2. Infinity and asymptotes
2.1. Limits to innity and vertical asymptotes. We have already discussed what it means to say
lim
xc
f(x) = +, but heres a friendly reminder. It means that as x comes closer and closer to c (from
either side), f(x) goes above every nite value and does not then come back down. You can similarly
understand what it means to say that lim
xc
f(x) = . You should also be able to understand the
one-sided versions of these concepts: lim
xc
f(x) = , lim
xc
+ f(x) = , lim
xc
f(x) = , and
lim
xc
+ f(x) = .
Now, heres a little guesswork question: if lim
xc
f(x) = , what can you say about lim
xc
f

(x)?
As a general rule, nothing, but in most of the situations that we see, it turns out that f

(x) also approaches

+
1
. The reason is easy to see graphically: for f(x) to head to + as x approaches a nite value, f needs
to climb faster and faster and faster.
Similarly, it is usually the case that if lim
xc
f(x) = , then lim
xc
f

(x) = . Also, if
lim
xc
+ f(x) = , then it is likely that lim
xc
+ f

(x) = (because it has to drop very quickly down

from innity) and if lim
xc
+ f(x) = , then lim
xc
+ f

(x) = (because it has to rise very quickly back

up from ). Again, these things are not always true, but for most of the typical examples youll see, they
will be.
Now heres the meaning of vertical asymptote. If lim
xc
f(x) = and/or lim
xc
+ f(x) = , then
the line x = c is termed a vertical asymptote for f. This is because the graph of f is approaching the vertical
line x = c. In some sense, if we think of f(c) = + or as the case may be, the vertical line becomes
the tangent line to the curve at that innite point.
Some of the typical situations worth noting are:
(1) lim
xc
f(x) = +from both sides. An example of this is the function f(x) = 1/x
2
with c = 0. The
vertical asymptote is the y-axis, i.e., the line x = 0. In this case, and in most other representative
examples, lim
xc
f

(x) = +, and lim

xc
+ f

(x) = .
(2) lim
xc
f(x) = from both sides. An example of this is the function f(x) = 1/x
2
with c = 0.
The vertical asymptote is the y-axis, i.e., the line x = 0. In this case, and in most other representative
examples, lim
xc
f(x) = , and lim
xc
+ f(x) = +.
1
More precisely, it turns out that if f

is continuous and does approach something, that something must be +. However,

there are weird examples where it oscillates
10
(3) lim
xc
f(x) = and lim
xc
+ f(x) = . Examples include f = tan at c = /2 (vertical
asymptote x = /2) and f(x) = 1/x at c = 0 (vertical asymptote x = 0). In both these cases, as
in most others, lim
xc
f

(x) = +.
11
(4) lim
xc
f(x) = and lim
xc
+ f(x) = . Examples include f = cot at c = 0 and f(x) := 1/x at
c = 0. In both these cases, as in most others, lim
xc
f

(x) = .
2.2. Horizontal asymptotes. Horizontal asymptotes are horizontal lines that the graph comes closer and
closer to, just as vertical asymptotes are vertical lines that the graph comes closer and closer to.
We saw that vertical asymptotes arose when the range value was approaching for a nite limiting
value of the domain. Horizontal asymptotes arise where the domain value approaches for a nite limiting
value of the range.
Explicitly, if lim
x
f(x) = L (with L a nite number), then the line y = L is a horizontal asymptote
for the graph of f, because as x , the graph comes closer and closer to this horizontal line. Similarly,
if lim
x
f(x) = M, then the line y = M is a horizontal asymptote for the graph of f. Thus, a function
whose domain extends to innity in both directions could have zero, one, or two horizontal asymptotes.
12
We will discuss some of the computational aspects of vertical and horizontal asymptotes in the problem
sessions. Later in the lecture, and in the addendum, we look at some computational tips and guidelines over
and above what is there in the book.
2.3. Vertical tangents. A vertical tangent to the graph of a function f occurs at a point (c, f(c)) if f is
continuous but not dierentiable at c, and lim
xc
f

(x) = + or lim
xc
f

(x) = . It is important that

the sign of innity in the limit is the same from both the left and the right side.
An example is the function f(x) := x
1/3
at the point c = 0. The function is continuous at 0. The
derivative functions is (1/3)x
2/3
, and the limit of this as x 0 (from either side) is +. Graphically,
what this means is that the tangent is vertical. In this case, the vertical tangent coincides with the y-axis,
because it is attained at the point 0.
13
Points of vertical tangent are points of inection, as we can see from the x
1/3
example. Recall that the
horizontal tangent case of the point of inection was typied by x
3
, and the general slogan was that the
function slows down for an instant to speed zero. For vertical tangents, we can think of it as the function
speeding up instantaneously to speed innity before returning to the realm of nite speed.
It is important to note that the situation of a vertical tangent requires that the function itself be dened
and continuous, and hence nite-valued, at the point. Thus, for instance, the function f(x) := 1/x satises
lim
x0
f

(x) = but does not have a vertical tangent at zero because the function is undened at zero.
2.4. Vertical cusps. A vertical cusp in the graph of f occurs at a point c if f is continuous at c, and both
one-sided limits of f

at c are innities of opposite sign. There are two possibilities:

(1) The left-hand limit of the derivative is + and the right-hand limit of the derivative is . Then,
(c, f(c)) is a point of local maximum. An example is f(x) := x
2/3
and c = 0 What happens in this
situation is the the graph has a sharp peak (picking up to speed innity) at the point c, after which
it rapidly starts dropping.
(2) The left-hand limit of the derivative is and the right-hand limit of the derivative is +. In this
case we get a local minimum. An example is f(x) := x
2/3
and c = 0.
There is a special curve called the astroid curve (I had planned to put this on the homework, but it went on
the chopping block when I needed to trim the homeworks to size), given by the equation x
2/3
+y
2/3
= a
2/3
.
This curve is not the graph of a function, since every value of x in (a, a) has two corresponding values of
y. Nonetheless, the curve is a good illustration of the concept of cusps: there are two vertical cusps at the
points (0, a) and (0, a) respectively, and two horizontal cusps at the points (a, 0) and (a, 0) respectively.
Shown below is the astroid curve for a = 1:
Note that for the graph of a function, the only kind of cusp that can occur is a vertical cusp, because
a horizontal or oblique cusp would result in the curve intersecting a vertical line at multiple points, which
would contradict the meaning of a function.
It is important to note that the situation of a vertical cusp requires that the function itself be dened
and continuous, and hence nite-valued, at the point. Thus, for instance, the function f(x) := 1/x
2
satises
lim
x0
f

(x) = + and lim

x0
+ f

(x) = but does not have a vertical tangent at zero because the
function is undened at zero.
3. Computational aspects
3.1. Computing limits at innity: a review. We review the main results that you have probably seen
and add some more:
(1) ()() =. In other words, if lim
xc
f(x) = and lim
xc
g(x) = , then lim
xc
f(x)g(x) =
. c could be nite or here, and we could take one-sided limits instead.
14
(2) ()() =, and ()() =.
(3) (a)() =if a > 0 and (a)() =if a < 0. Similarly, (a)() =
if a > 0 and (a)() = if a < 0.
(4) The previous point can be generalized somewhat: ( ), times a function that eventually has a
positive lower bound (even if it keeps oscillating), is also . Analogous results hold for negative
upper bounds.
(5) (0)() is an indeterminate form: it is not clear what it tends to without doing more work.
(6) () + () =.
(7) () + (a) = where a is nite. More generally, plus anything that is bounded from
below is also .
(8) () () and () + () are indeterminate forms.
Apart from this, the main facts you need to remember are that if a > 0, then lim
x
x
a
= and
lim
x
x
a
= lim
x
x
a
= 0. Note that this holds regardless of whether a is an integer.
When a is an odd integer or a rational number with odd numerator and odd denominator, lim
x
x
a
=
. When a is an even integer or a rational number with even numerator and odd denominator, lim
x
x
a
=
.
Also worth noting: lim
x0
+ x
a
= for a > 0 and lim
x0
x
a
= lim
x
x
a
, which is computed by
the rule above.
We can use these facts to explain most of the limits involving polynomial and rational functions. Earlier,
we had noted that when calculating the limit of a polynomial, it is enough to calculate the limit of its leading
monomial. Lets now see why.
Consider the function f(x) := x
7
5x
5
+3x+2. Then, we can write f(x) = x
7

1 5x
2
+ 3x
6
+ 2x
7

.
The expression on the inside is 1 plus various negative powers of x. Each of those negative powers of x goes
to 0 as x . So, we obtain:
lim
x
[1 5x
2
+ 3x
6
+ 2x
7
] = 1
We also have lim
x
x
7
= . Thus, the limit of the product is .
Lets now consider an example of a rational function:
9x
3
3x + 2
103x
2
17x 99
Earlier, we had discussed that when computing such limits at , we can simply calculate thel imits of
the leading terms and ignore the rest. We now have a better understanding of the rationale behind this.
Formally:
lim
x
9x
3
3x + 2
103x
2
17x 99
= lim
x
x
3
(9 3x
2
+ 2x
3
)
x
2
(103 17x
1
99x
2
)
=limx x lim
x
9 3x
2
+ 2x
3
103 17x
1
99x
2
= lim
x
x
9
103
= +
More generally, we see that if the degree of the numerator is greater than the degree of the denominator,
the fraction approaches as x , with the sign depending on the signs of the leading coecients
and the parity (even versus odd) of the exponents.
If the numerator and denominator have equal degree, the limit is a nite number. For x , it is
the ratio of the leading coecients (notice that it is the same on both sides). This is the case where we get
horizontal asymptotes. In this case, the horizontal asymptotes on both ends coincide.
For instance:
15
lim
x
2x
2
3x + 5
23x
2
x 1
= lim
x
x
2
(2 3x
1
+ 5x
2
)
x
2
(23 x
1
x
2
)
= lim
x
x
2
x
2
lim
x
2 3x
1
+ 5x
2
23 x
1
x
2
=1
2
23
=
2
23
Finally, when the degree of the numerator is less than the degree of the denominator, then the fraction
tends to 0 as x and also tends to 0 as x . Thus, in this case, we get the x-axis as the horizontal
asymptote on both sides.
3.2. The 1/x substitution trick.
Consider the limit:
lim
x
xsin(1/x)
This limit cannot be computed by plugging in values, because x and 1/x 0, so sin(1/x) 0,
and we get the indeterminate form ( )( 0). The approach we use here is to set t = 1/x. As x ,
t 0
+
. Since t = 1/x, we get x = 1/t. Plugging in, we get:
lim
t0
+
sint
t
This limit is 1, as we know well.
Note that with this general substitution, limits to innity correspond to right-hand limits at 0 for the
reciprocal and limits at correspond to left-hand limits at 0 for the reciprocal. If there is a two-sided
limit at 0 for the reciprocal, the limits at are the same. In fact, in the xsin(1/x) example, the limits at
and are both 1 since lim
t0
sint/t = 1.
3.3. Dierence of square roots. Consider the limit:
lim
x
(

x + 1

x)
There are many ways to compute this limit, but the easiest is to use the general 1/x substitution trick.
Let t = 1/x. Then the above limit becomes:
lim
t0
+

t + 1 1

t
This is an indeterminate form (specically, a 0/0 form). However, we can do the rationalization trick and
rewrite this as:
16
lim
t0
+
t

t(

t + 1 + 1)
The t and

t cancel to give a

t in the numerator, and we can evaluate and nd the limit to be 0.
A similar approach can be used to handle, for instance, a dierence of cube roots.
3.4. Combinations of polynomial and trigonometric functions. We illustrate using some examples:
(1) Consider the function f(x) := x + 25 sinx. As x , this is the sum of a function that tends to
innity and a function that oscillates. The oscillating component, however, has a nite lower bound,
and hence, lim
x
f(x) = . Similarly, lim
x
f(x) = .
(2) Consider the function f(x) := xsinx. As x , this is the product of a function that goes to
and a function that oscillates between 1 and 1. The oscillating part causes the sign of the
whole expression to shift, and so as x , f(x) is oscillating with an ever-increasing magnitude of
oscillation. A similar observation holds for x .
(3) Consider the function f(x) := x(3 + sinx) As x , this is the product of a function that tends
to and a function that oscillates between 2 and 4. The important point here is that the latter
oscillation has a positive lower bound, so the product still tends to .
(4) Consider the function f(x) := xsin(1/x). As x , this is the product of a function that tends to
and a function that tends to 0, so it is an indeterminate form. We already discussed above how
this particular indeterminate form can be handled.
17